Toward performance-portable PETSc for GPU-based exascale systems

The Portable Extensible Toolkit for Scientific computation (PETSc) library delivers scalable solvers nonlinear time-dependent differential and algebraic equations numerical optimization. PETSc design performance portability addresses fundamental GPU accelerator challenges stresses flexibility extensibility by separating the programming model used application from that library, it enables developers to use their preferred model, such as Kokkos, RAJA, SYCL, HIP, CUDA, or OpenCL, on upcoming exascale systems. A blueprint using GPUs PETSc-based codes is provided, case studies emphasize high achieved current GPU-based